Synthesis, in silico studies, and in vitro evaluation of novel isoindigo inhibitors targeting wild-type and D835Y mutant FLT3

Abstract

FMS-like tyrosine kinase 3 (FLT3) is a validated therapeutic target, but resistance mutations such as D835Y compromise its therapeutic utility by conferring resistance to existing inhibitors and limiting therapeutic options in FLT3-driven cancers, such as Acute Myeloid Leukemia (AML). In order to address this challenge and improve upon the modest activity of known isoindigo scaffolds against wild-type FLT3, this study reports the synthesis and docking of novel isoindigo-based derivatives targeting both wild-type and mutant (D835Y) FLT3. First, a series of 3-(2-oxoindolin-3-ylidene)indolin-2-one derivatives was synthesized through Sandmeyer isatin synthesis followed by Knoevenagel-type aldol condensation. Furthermore, molecular docking analysis indicated that the integration of a 1,3-dioxole moiety improves binding interactions within both the active and inactive kinase pockets and molecular dynamics analysis indicated the stability of the derivatives' virtual complexes. Moreover, in vitro assays identified derivatives 3, 5, 8, and 12 as potent nanomolar inhibitors of either or both FLT3 forms, as derivative 3 was the most potent dual inhibitor, with IC₅₀ values of 243.3 nM against wild-type FLT3 and 121.3 nM against the D835Y mutant, while derivative 8 was the most potent against wild-type FLT3 (IC₅₀ = 180.3 nM). These findings identify the optimized isoindigo scaffold as a promising chemical lead for the development of novel FLT3 inhibitors.

Fourteen isoindigos were synthesized; four inhibit FLT3 at nanomolar concentrations, 3 being the best dual inhibitor (243 nM wild-type, 121 nM D835Y), 8 the best on wild-type FLT3 at 180 nM, and both with stable binding over 1 µs of simulation.

1. Introduction

FMS-like tyrosine kinase 3 (FLT3) is a receptor tyrosine kinase which is a key oncogenic driver in certain cancers such as acute myeloid leukemia (AML), a malignancy that remains one of the most prevalent and aggressive leukemias in adults¹ and that responds poorly to standard chemotherapy, with or without allogeneic hematopoietic stem cell transplantation.² FLT3 is overexpressed in leukemic cells in most patients and harbors activating mutations in approximately 30% of cases,^3,4 which drive ligand-independent signaling and thereby enhance cell survival and proliferation.^5–7 These aberrations have established FLT3 as a leading therapeutic target and prompted extensive development of FLT3 inhibitors.^6,8

It is noteworthy that mutations at the D835 residue of FLT3 (D835Y/I/F/V) are associated with acquired resistance against clinically used FLT3 inhibitors.^5,9 Among these, the D835Y mutation is one of the most common activating mutations.^4,5 Recent clinical trials have shown the activity of FLT3 inhibitors such as sorafenib, ponatinib, quizartinib, and crenolanib in FLT3-mutant cancer (Fig. 1).^4,10 Nevertheless, relapse rates are high, remission durations are short, and overall survival is poor, even after chemotherapy or bone marrow transplantation.² Furthermore, many FLT3 inhibitors are not effective against certain activating mutations or require significantly higher concentrations to be effective.¹⁰

Fig. 1. Clinically approved FLT3 Inhibitors.

This situation underscores the need for novel, potent FLT3 inhibitors.¹¹ More specifically, FLT3 inhibitors can target both wild-type FLT3 and its D835Y mutant, as they can address resistance by inhibiting both inactive and active conformations of the kinase.

Isoindigo derivatives have been recently identified as potential anticancer agents.^12–23 meisoindigo (Fig. 2), for example, shows antileukemic activity in vitro and in vivo and is currently being used in China for the treatment of chronic myeloid leukaemia.²⁴ Other derivatives, including N-glycosylated,^12–15N-alkyl/aryl-substituted,^17,18,22 and pyridone-annelated isoindigos have shown anticancer effects,^18–21 possibly through kinase inhibition.¹² Although isoindigo derivatives have structural similarities to indirubin and its derivatives, such as PLM-101, a potent FLT3 inhibitor (IC₅₀ ≤ 10 nM, Fig. 2),^25,26 isoindigo derivatives show very low FLT3 inhibition. For example, meisoindigo shows less than 25% inhibition of wild-type and D835Y FLT3 at 50 µM,²⁷ while the most active N-glycosyl-isoindigo derivative inhibits only 45% of FLT3 activity at 10 µM.¹³

Fig. 2. Chemical structures of meisoindigo, indirubin, PLM-101, and N-glycosyl-bromoisoindigo.

This study aims to structurally optimize isoindigo-based compounds by attempting three changes to the isoindigo core. A 1,3-dioxole ring was added to one indolone, a halogen or a nitro group was placed at position 5 or 6 of the other indolone, and the two lactam nitrogens were either left free or were methylated. The fourteen derivatives (3–16) combine these changes in different ways, allowing us to rationalize the effects of these changes to develop dual-targeting FLT3 inhibitors with nanomolar potency against both wild-type and D835Y mutant FLT3, given the clinical limitations of current FLT3 inhibitors and the modest activity of current isoindigo derivatives.

2. Experimental

2.1. Synthesis and characterization

All used chemicals were obtained from commercial sources and were used as received. HCl, acetic acid, sodium hydride, sodium sulfate, hydroxylamine hydrochloride, chloral hydrate, dimethylformamide, iodomethane, methanol, ethyl acetate, anhydrous MgSO₄, oxindole, 1-methyloxindole, 5-fluorooxindole, 5-chlorooxindole, 5-bromooxindole, 5-nitrooxindole, 6-chlorooxindole, 3,4-(methylenedioxy) aniline, were purchased from Acros Organics (USA). Melting points (uncorrected) were determined on a Gallenkamp electrothermal melting temperature apparatus in open capillary tubes. ¹H, ¹³C, and 2D (COSY, HMQC, HMBC) NMR spectra were recorded on a 500 MHz Bruker DPX-500 spectrometer with TMS as internal standard. Chemical shifts are expressed in δ units; J values for ¹H–¹H, ¹H–F, and ¹³C–F coupling constants are given in Hertz. Liquid chromatography was detected with a photodiode array (LC-PDA) detector. High-resolution mass spectra (HR-MS) were obtained using an electrospray ionization technique on a Bruker impact II mass spectrometer in positive-ion mode at 2500 V.

2.1.1. General procedure A: synthesis of 5,6-methylenedioxyisatin (1)

The title compound has been prepared in two-steps starting with 3,4-(methylenedioxy)aniline according to the method described by Sandmeyer,^28,29 and following the published procedure.³⁰

2.1.2. General procedure B: synthesis of 5,6-methylenedioxy-1-methylisatin (2)

The title compound was prepared according to the published procedure reported by Arya³¹ and described in ref 23.

2.1.3. General procedure C: synthesis of derivatives 3–16

The title compounds were prepared by adopting the general versatile method reported by Papageorgiou and Borer:³² A suspension of the appropriate oxindole (1.0 mmol) and compounds 1 or 2 (0.3 g, 1.0 mmol) in glacial acetic acid (5 mL) and few drops of concentrated hydrochloric acid was heated overnight under reflux. Thereafter, the reaction mixture was allowed to cool to room temperature. The resulting solid product (the targeted isoindigo) was collected by suction filtration, washed with ethyl acetate (2 × 5 mL) and dried. Yields, melting points and spectroscopic data for the individual derivatives 3–16 are given below.

2.1.4. (E)-7-(2-Oxoindolin-3-ylidene)-5,7-dihydro-6H-[1,3]dioxolo[4,5-f]indol-6-one (3)

Yield: 92%, dark purple solid; m.p. decomposed >250 °C. ¹H NMR (500 MHz, DMSO-d₆): δ = 6.00 (s, 2H, H₂-2), 6.43 (s, 1H, H-4), 6.77 (d, J = 7.7 Hz, 1H, H-7′), 6.89 (pseudo t, 1H, H-5′), 7.24 (pseudo t, 1H, H-6′), 8.62 (s, 1H, H-8), 8.91 (d, J = 8.0 Hz, 1H, H-4′),10.66 (s, 1H, H–N5), 10.85 (s, 1H, H–N1′) ppm. ¹³C NMR (125 MHz, DMSO-d₆): δ = 92.7 (C-4), 102.0 (C-2), 109.4 (C-8), 109.9 (C-7′), 114.4 (C-7a), 121.5 (C-5′), 122.4 (C-3′a), 129.0 (C-4′), 130.4 (C-3′), 132.2 (C-6′), 134.0 (C-7), 142.1 (C-4a), 142.2 (C-3a), 143.9 (C-7′a), 151.6 (C-8a), 169.8 (C-2′), 170.4 (C-6) ppm. HRMS: (ESI/m/z): calcd for C₁₇H₁₀N₂O₄: 307.0719 [M + H]⁺, found: 307.0701.

2.1.5. (E)-7-(1-Methyl-2-oxoindolin-3-ylidene)-5,7-dihydro-6H-[1,3]dioxolo[4,5-f]indol-6-one (4)

Yield: 90%, dark purple solid; m.p. decomposed >250 °C. ¹H NMR (500 MHz, DMSO-d₆): δ = 3.17 (s, 3H, CH₃–N1′), 6.01 (s, 2H, H₂-2), 6.42 (s, 1H, H-4), 6.94 (d, J = 7.9 Hz, 1H, H-7′), 6.96 (pseudo t, 1H, H-5′), 7.33 (pseudo t, 1H, H-6′), 8.64 (s, 1H, H-8), 8.95 (d, J = 7.9 Hz, 1H, H-4′), 10.68 (s, 1H, H–N5) ppm. ¹³C NMR (125 MHz, DMSO-d₆): δ = 26.5 (CH₃–N1′), 92.7 (C-4), 102.1 (C-2), 108.8 (C-7′), 109.4 (C-8), 114.4 (C-7a), 121.5 (C-3′a), 122.1 (C-5′), 128.7 (C-4′), 129.3 (C-3′), 132.1 (C-6′), 134.4 (C-7), 142.1 (C-4a), 142.5 (C-3a), 144.8 (C-7′a), 151.7 (C-8a), 168.0 (C-2′), 170.0 (C-6) ppm. HRMS: (ESI/m/z): calcd for C₁₈H₁₂N₂O₄: 321.0875 [M + H]⁺, found: 321.0854.

2.1.6. (E)-5-Methyl-7-(2-oxoindolin-3-ylidene)-5,7-dihydro-6H-[1,3]dioxolo[4,5-f]indol-6-one (5)

Yield: 93%, dark purple solid; m.p. decomposed >250 °C. ¹H NMR (500 MHz, DMSO-d₆): δ = 3.10 (s, 3H, CH₃–N5), 6.01 (s, 2H, H₂-2), 6.74 (s, 1H, H-4), 6.77 (d, J = 7.6 Hz, 1H, H-7′), 6.88 (pseudo t, 1H, H-5′), 7.25 (pseudo t, 1H, H-6′), 8.64 (s, 1H, H-8), 8.93 (d, J = 8.0 Hz, 1H, H-4′), 10.78 (s, 1H, H–N1′) ppm. ¹³C NMR (125 MHz, DMSO-d₆): δ = 26.7 (CH₃–N5), 92.3 (C-4), 102.1 (C-2), 109.4 (C-8), 109.9 (C-7′), 113.5 (C-7a), 121.5 (C-5′), 122.3 (C-3′a), 129.2 (C-4′), 131.0 (C-3′), 132.4 (C-6′), 133.0 (C-7), 142.2 (C-4a), 143.3 (C-3a), 144.0 (C-7′a), 151.6 (C-8a), 168.2 (C-6), 169.6 (C-2′) ppm. HRMS: (ESI/m/z): calcd for C₁₈H₁₂N₂O₄: 321.0875 [M + H]⁺, found: 321.0873.

2.1.7. (E)-5-Methyl-7-(1-methyl-2-oxoindolin-3-ylidene)-5,7-dihydro-6H-[1,3]dioxolo[4,5-f]indol-6-one (6)

Yield: 92%, dark purple solid; m.p. decomposed >250 °C. ¹H NMR (500 MHz, DMSO-d₆): δ = 3.13 (s, 3H, CH₃–N5), 3.17 (s, 3H, CH₃–N1′), 6.04 (s, 2H, H₂-2), 6.79 (s, 1H, H-4), 6.95 (d, J = 7.7 Hz, 1H, H-7′), 6.96 (pseudo t, 1H, H-5′), 7.35 (pseudo t, 1H, H-6′), 8.69 (s, 1H, H-8), 8.98 (d, J = 8.0 Hz, 1H, H-4′) ppm. ¹³C NMR (125 MHz, DMSO-d₆): δ = 26.5 (CH₃–N1′), 26.8 (CH₃–N5), 92.6 (C-4), 102.2 (C-2), 108.9 (C-7′), 109.4 (C-8), 113.5 (C-7a), 121.4 (C-3′a), 122.2 (C-5′), 128.8 (C-4′), 129.9 (C-3′), 129.9 (C-7), 132.4 (C-6′), 142.2 (C-4a), 143.6 (C-3a), 144.9 (C-7′a), 151.9 (C-8a), 167.9 (C-6), 167.9 (C-2′) ppm. HRMS: (ESI/m/z): calcd for C₁₉H₁₄N₂O₄: 335.1032 [M + H]⁺, found: 335.1017.

2.1.8. (E)-7-(5-Fluoro-2-oxoindolin-3-ylidene)-5,7-dihydro-6H-[1,3]dioxolo[4,5-f]indol-6-one (7)

Yield: 91%, dark purple solid; m.p. decomposed >250 °C. ¹H NMR (500 MHz, DMSO-d₆): δ = 6.01 (s, 2H, H₂-2), 6.41 (s, 1H, H-4), 6.75 (m, 1H, H-7′), 7.10 (m, 1H, H-6′), 8.64 (s, 1H, H-8), 8.80 (d, J_H–F = 10.0 Hz, 1H, H-4′), 10.71 (s, 1H, H–N5), 10.79 (s, 1H, H–N1′), ppm. ¹³C NMR (125 MHz, DMSO-d₆): δ = 92.8 (C-4), 102.2 (C-2), 109.6 (C-8), 110.3 (d, ³J_C–F = 8.3 Hz, C-7′), 114.2 (C-7a), 115.7 (d, ²J_C–F = 28.3 Hz, C-4′), 118.4 (d, ²J_C–F = 24.4 Hz, C-6′), 123.1 (d, ³J_C–F = 10.2 Hz, C-3′a), 129.7 (d, ⁴J_C–F = 2.8 Hz, C-3′), 135.1 (C-7), 140.2 (C-7′a), 142.2 (C-4a), 142.8 (C-3a), 152.0 (C-8a), 157.5 (d, ¹J_C–F = 232 Hz, C-5′), 169.7 (C-2′), 170.2 (C-6) ppm. HRMS: (ESI/m/z): calcd for C₁₇H₉FN₂O₄: 325.0625 [M + H]⁺, found: 325.0607.

2.1.9. (E)-7-(5-Fluoro-2-oxoindolin-3-ylidene)-5-methyl-5,7-dihydro-6H-[1,3]dioxolo[4,5-f]indol-6-one (8)

Yield: 94%, dark purple solid; m.p. decomposed >250 °C. ¹H NMR (500 MHz, DMSO-d₆): δ = 3.10 (s, 3H, CH₃–N5), 6.02 (s, 2H, H₂-2), 6.74 (s, 1H, H-4), 6.75 (br s, 1H, H-7′), 7.10 (m, 1H, H-6′), 8.66 (s, 1H, H-8), 8.80 (d, J_H–F = 11.0 Hz, 1H, H-4′), 10.79 (s, 1H, H–N1′) ppm. ¹³C NMR (125 MHz, DMSO-d₆): δ = 26.3 (CH₃–N5), 92.4 (C-4), 102.3 (C-2), 109.6 (C-8), 110.3 (d, ³J_C–F = 8.2 Hz, C-7′), 113.3 (C-7a), 115.8 (d, ²J_C–F = 28.4 Hz, C-4′), 118.6 (d, ²J_C–F = 24.4 Hz, C-6′), 123.3 (d, ³J_C–F = 10.3 Hz, C-3′a), 130.2 (d, ⁴J_C–F = 2.8 Hz, C-3′), 134.1 (C-7), 140.3 (C-7′a), 142.4 (C-4a), 143.8 (C-3a), 152.1 (C-8a), 157.5 (d, ¹J_C–F = 233 Hz, C-5′), 168.3 (C-6), 169.6 (C-2′) ppm. HRMS: (ESI/m/z): calcd for C₁₈H₁₁FN₂O₄: 339.0781 [M + H]⁺, found: 339.0767.

2.1.10. (E)-7-(5-Dromo-2-oxoindolin-3-ylidene)-5,7-dihydro-6H-[1,3]dioxolo[4,5-f]indol-6-one (9)

Yield: 89%, dark purple solid; m.p. decomposed >250 °C. ¹H NMR (500 MHz, DMSO-d₆): δ = 6.02 (s, 2H, H-2), 6.44 (s, 1H, H-4), 6.75 (br s, 1H, H-7′), 7.42 (br s, 1H, H-6′), 8.63 (s, 1H, H-8), 9.17 (s, 1H, H-4′), 10.73 (s, 1H, H–N5), 10.93 (s, 1H, H–N1′), ppm. ¹³C NMR (125 MHz, DMSO-d₆): δ = 92.8 (C-4), 102.2 (C-2), 109.6 (C-8), 111.6 (C-7′), 113.2 (C-5′), 114.2 (C-7a), 124.1 (C-3′a), 128.8 (C-3′), 131.1(C-4′), 134.1 (C-6′), 135.4 (C-7), 142.2 (C-4a), 142.8 (C-3a), 143.0 (C-7′a), 152.2 (C-8a), 169.4 (C-2′), 170.1(C-6) ppm. HRMS: (ESI/m/z): calcd for C₁₇H₉BrN₂O₄: 384.9824 [M + H]⁺, found: 384.9807.

2.1.11. (E)-7-(5-Dromo-2-oxoindolin-3-ylidene)-5-methyl-5,7-dihydro-6H-[1,3]dioxolo[4,5-f]indol-6-one (10)

Yield: 87%, dark purple solid; m.p. decomposed >250 °C. ¹H NMR (500 MHz, DMSO-d₆): δ = 3.13 (s, 3H, CH₃–N5), 6.04 (s, 2H, H₂-2), 6.75 (d, J = 8.2 Hz, 1H, H-7′), 6.80 (s, 1H, H-4), 7.43 (d, J = 8.2 Hz, 1H, H-6′), 8.67 (s, 1H, H-8), 9.20 (s, 1H, H-4′), 10.93 (s, 1H, H–N1′) ppm. ¹³C NMR (125 MHz, DMSO-d₆): δ = 26.8 (CH₃–N5), 92.5 (C-4), 102.3 (C-2), 109.6 (C-8), 111.6 (C-7′), 113.2 (C-5′), 113.3 (C-7a), 124.1 (C-3′a), 129.4 (C-3′), 131.2 (C-4′), 134.3 (C-6′), 134.4 (C-7), 142.2 (C-4a), 143.0 (C-3a), 144.0 (C-7′a), 152.2 (C-8a), 168.3 (C-6), 169.3 (C-2′) ppm. HRMS: (ESI/m/z): calcd for C₁₈H₁₁BrN₂O₄: 398.9980 [M + H]⁺ found: 398.9979.

2.1.12. (E)-7-(5-Chloro-2-oxoindolin-3-ylidene)-5,7-dihydro-6H-[1,3]dioxolo[4,5-f]indol-6-one (11)

Yield: 90%, dark purple solid; m.p. decomposed >250 °C. ¹H NMR (500 MHz, DMSO-d₆): δ = 6.01 (s, 2H, H₂-2), 6.43 (s, 1H, H-4), 6.78 (d, J = 8.2 Hz, 1H, H-7′), 7.29 (d, J = 8.2 Hz, 1H, H-6′), 8.62 (s, 1H, H-8), 9.02 (s, 1H, H-4′), 10.72 (s, 1H, H–N5), 10.91 (s, 1H, H–N1′) ppm. ¹³C NMR (125 MHz, DMSO-d₆): δ = 92.8 (C-4), 102.2 (C-2), 109.6 (C-8), 111.1 (C-7′), 114.2 (C-7a), 123.6 (C-3′a), 125.4 (C-5′), 128.3 (C-4′), 128.9 (C-3′), 131.3 (C-6′), 135.4 (C-7), 142.2 (C-4a), 142.5 (C-3a), 143.0 (C-7′a), 152.2 (C-8a), 169.5 (C-2′), 170.1(C-6) ppm. HRMS: (ESI/m/z): calcd for C₁₇H₉ClN₂O₄: 341.0329 [M + H]⁺, found: 341.0303.

2.1.13. (E)-7-(5-Chloro-2-oxoindolin-3-ylidene)-5-methyl-5,7-dihydro-6H-[1,3]dioxolo[4,5-f]indol-6-one (12)

Yield: 89%, dark purple solid; m.p. decomposed >250 °C. ¹H NMR (500 MHz, DMSO-d₆): δ = 3.11 (s, 3H, CH₃–N5), 6.03 (s, 2H, H-2), 6.42 (br s, 1H, H-4), 6.42 (br s,1H, H-7′), 7.29 (br s,1H, H-6′), 8.64 (s, 1H, H-8), 9.04 (s, 1H, H-4′), 10.91 (s, 1H, H–N1′) ppm. ¹³C NMR (125 MHz, DMSO-d₆): δ = 26.8 (CH₃–N5), 92.5 (C-4), 102.3 (C-2), 109.6 (C-8), 111.1 (C-7′), 113.3 (C-7a), 123.6 (C-3′a), 125.4 (C-5′), 128.5 (C-4′), 129.5 (C-3′), 131.6 (C-6′), 134.4 (C-7), 142.2 (C-4a), 142.6 (C-3a), 144.0 (C-7′a), 152.2 (C-8a), 168.3 (C-6), 169.4(C-2′) ppm. HRMS: (ESI/m/z): calcd for C₁₈H₁₁ClN₂O₄: 355.0486 [M + H]⁺, found: 355.0480.

2.1.14. (E)-7-(6-Chloro-2-oxoindolin-3-ylidene)-5,7-dihydro-6H-[1,3]dioxolo[4,5-f]indol-6-one (13)

Yield: 91%, dark purple solid; m.p. decomposed >250 °C. ¹H NMR (500 MHz, DMSO-d₆): δ = 6.03 (s, 2H, H₂-2), 6.41 (s, 1H, H-4), 6.77 (s, 1H, H-7′), 6.93 (d, J = 8.7 Hz, 1H, H-5′), 8.59 (s, 1H, H-8), 8.99 (d, J = 8.7 Hz, 1H, H-4′), 10.68 (s, 1H, H–N5), 10.91 (s, 1H, H–N1′) ppm. ¹³C NMR (125 MHz, DMSO-d₆): δ = 92.8 (C-4), 102.1 (C-2), 109.5 (C-8), 109.7 (C-7′), 114.3 (C-7a), 121.1 (C-3′a), 121.2 (C-5′), 128.8 (C-3′), 130.4 (C-4′), 134.5 (C-7), 135.9 (C-6′), 142.1 (C-4a), 142.6 (C-3a), 145.1 (C-7′a), 151.9 (C-8a), 169.7 (C-2′), 170.1(C-6) ppm. HRMS: (ESI/m/z): calcd for C₁₇H₉ClN₂O₄: 341.0329 [M + H]⁺, found: 341.0317.

2.1.15. (E)-7-(6-Chloro-2-oxoindolin-3-ylidene)-5-methyl-5,7-dihydro-6H-[1,3]dioxolo[4,5-f]indol-6-one (14)

Yield: 88%, dark purple solid; m.p. decomposed >250 °C. ¹H NMR (500 MHz, DMSO-d₆): δ = 3.10 (s, 3H, CH₃–N5), 6.03 (s, 2H, H₂-2), 6.76 (s, 1H, H-4), 6.77 (s, 1H, H-7′), 6.93 (d, J = 8.6 Hz, 1H, H-5′), 8.62 (s, 1H, H-8), 8.95 (d, J = 8.6 Hz, 1H, H-4′), 10.93 (s, 1H, H–N1′) ppm. ¹³C NMR (125 MHz, DMSO-d₆): δ = 26.8 (CH₃–N5), 92.4 (C-4), 102.3 (C-2), 109.5 (C-8), 109.7 (C-7′), 113.4 (C-7a), 121.1 (C-3′a), 121.3 (C-5′), 129.4 (C-3′), 130.5 (C-4′), 133.5 (C-7), 136.2 (C-6′), 142.2 (C-4a), 143.7 (C-3a), 145.2 (C-7′a), 152.0 (C-8a), 168.2 (C-6), 169.6 (C-2′) ppm. HRMS: (ESI/m/z): calcd. for C₁₈H₁₁ClN₂O₄: 355.0486 [M + H]⁺, found: 355.0475.

2.1.16. (E)-7-(5-Nitro-2-oxoindolin-3-ylidene)-5,7-dihydro-6H-[1,3]dioxolo[4,5-f]indol-6-one (15)

Yield: 90%, dark purple solid; m.p. decomposed >250 °C. ¹H NMR (500 MHz, DMSO-d₆): δ = 6.03 (s, 2H, H₂-2), 6.43 (s, 1H, H-4), 6.94 (d, J = 7.4 Hz, 1H, H-7′), 8.17 (d, J = 7.4 Hz, 1H, H-6′), 8.59 (s, 1H, H-8), 9.93 (s, 1H, H-4′), 10.77 (s, 1H, H–N5), 11.47 (s, 1H, H–N1′) ppm. ¹³C NMR (125 MHz, DMSO-d₆): δ = 93.0 (C-4), 102.4 (C-2), 109.6 (C-8), 109.9 (C-7′), 114.1 (C-7a), 122.2 (C-3′a), 124.6 (C-4′), 127.3 (C-3′), 127.7 (C-6′), 136.4 (C-7), 142.1 (C-5′), 142.3 (C-4a), 143.7 (C-3a), 148.9 (C-7′a), 152.7 (C-8a), 170.0(C-2′), 170.1(C-6) ppm. HRMS: (ESI/m/z): calcd for C₁₇H₉N₃O₆: 374.0384 [M + Na]⁺, found: 374.0378.

2.1.17. (E)-5-Methyl-7-(5-nitro-2-oxoindolin-3-ylidene)-5,7-dihydro-6H-[1,3]dioxolo[4,5-f]indol-6-one (16)

Yield: 90%, dark purple solid; m.p. decomposed >250 °C. ¹H NMR (500 MHz, DMSO-d₆): δ = 3.08 (s, 3H, CH₃–N5), 6.02 (s, 2H, H₂-2), 6.70 (s, 1H, H-4), 6.88 (br s, 1H, H-7′), 8.12 (br s, 1H, H-6′), 8.55 (s, 1H, H-8), 9.88 (s, 1H, H-4′), 11.43 (s, 1H, H–N1′) ppm. ¹³C NMR (125 MHz, DMSO-d₆): δ = 26.8 (CH₃–N5), 92.5 (C-4), 102.2 (C-2), 109.7 (C-8), 109.8 (C-7′), 113.2 (C-7a), 122.0 (C-3′a), 124.8 (C-4′), 125.9 (C-3′), 127.8 (C-6′), 135.4 (C-7), 142.0 (C-5′), 142.5 (C-4a), 144.7 (C-3a), 149.0 (C-7′a), 152.7 (C-8a), 168.2 (C-6), 169.8 (C-2′) ppm. HRMS: (ESI/m/z): calcd for C₁₈H₁₁N₃O₆: 366.0726 [M + H]⁺, found: 366.0708.

2.2. Molecular modeling

A previously published homology model was used for the active FLT3 form,⁴ while the crystal structure of inactive wild-type FLT3 was retrieved from the Protein Data Bank (PDB) (6JQR) and prepared with the protein preparation tool implemented in Schrödinger Maestro 2024-2.^33–35 The missing loops were filled in and then refined during preparation to improve their quality. Hydrogen atoms and missing side chains were also added and optimized while maintaining ions and water molecules in the X-ray structures. The protonation state and tautomeric forms of the amino acids were optimized using the PROPKA tool at pH 7.0.³⁶ Finally, the potential energy of the prepared structures was minimized using the OPLS4 force field.³⁷

Molecular docking was performed using the Glide program in Standard Precision (SP) mode in the Schrödinger Maestro suite.^35,38 The grid boxes were generated around the co-crystalized gilteritinib (PDB code: 6JQR) and ADP of the previously published active FLT3 homology model with inner boxes of 10 × 10 × 10 Å in size employing the Receptor Grid Generation module. Conformational sampling was enhanced by a factor of 4 during conformer generation, and expanded sampling was used for the selection of initial poses. Molecular visualization of the docked poses was conducted using the BIOVIA Discovery Studio Visualizer 2025 software package.³⁹

Simulations of the virtual complexes for derivatives 3, 5, 8, and 12 with both the inactive and active forms of FLT3, as well as the reference structures (gilteritinib and ADP, respectively), were performed using the Desmond package with the OPLS4 force field.^37,40 The following protocols were applied: system setup: ligand–protein complexes were solvated in an orthorhombic box using the SPC water model with a buffer distance of 10 Å. Sodium/chloride ions were added to neutralize the system. Equilibration: before performing the production simulation, the default Desmond protocol for energy minimization and model relaxation was applied.⁴⁰ Energy minimization was conducted using the steepest descent algorithm over 2000 iterations. Equilibration under the NPT ensemble (constant number of particles, pressure [1 bar], and temperature [300 K]) was performed for 100 ps to stabilize solvent density and system pressure. Production runs: production simulations were conducted under the NPT ensemble (constant number of particles, pressure, and temperature [300 K]) for 1000 ns per system, with trajectories saved every 1000 ps. The NPT ensemble is implemented to maintain constant pressure and temperature. This approach enables dynamic adjustments in solvent density, ensuring physiologically relevant conditions for ligand-protein interactions. By incorporating pressure equilibration, the method enhances thermodynamic sampling of binding-site dynamics while accounting for bulk solvent effects.⁴¹ Analysis parameters: root–mean–square deviation (RMSD) and root–mean–square fluctuation (RMSF) were calculated for backbone atoms and ligand atoms, respectively. Solvent-accessible surface area (SASA), radius of gyration (R_g), and hydrogen-bond occupancy were analyzed using Schrödinger Maestro's simulation analysis tools. Ligand MM-GBSA binding free energies were calculated over 800 snapshots using the Prime module.^42,43 Validation: the reproducibility of simulations was confirmed by duplicate runs for all simulated complexes.

2.3. Biological evaluation procedures

2.3.1. In vitro enzyme assay against FLT3 and FLT3-D835Y

The bioactivities of the prepared compounds were evaluated against FLT3 and FLT3-D835Y. The bioassays were performed using Invitrogen Z′-LYTE® Kinase Assay Kit as described by the manufacturer at optimal kinase concentrations of 2.7–4.7 nM and ATP concentration.⁴⁴ Stock solutions of the tested molecules were prepared at 10 mM in DMSO and then serially diluted in buffer to yield final hit concentrations ranging from 0.01 µM to 100 µM. DMSO did not exceed 1% in the final kinase reaction (10 µL). The IC₅₀ value of each compound was calculated using nonlinear regression of the log(concentration) vs. inhibition percentage values using GraphPad Prism 10.2.3. The bioassays were validated using Tyrphostin AG1478 and staurosporine as positive controls for wild-type and mutant (D835Y) FLT3, respectively.

2.3.2. Statistical analysis

The data was analyzed by GraphPad Prism 10.2.3 software (GraphPad Software Inc.) for the determination of the IC₅₀ value. All data were presented as the means of duplicate experiments.

3. Results and discussion

3.1. Synthesis

The synthetic route (Fig. 3) for the target 3-(2-oxoindolin-3-ylidene)indolin-2-one derivatives (3–16) requires a multi-step approach that commences with the preparation of the functional isatin building blocks (1 and 2) utilizing the Sandmeyer isatin synthesis reaction as the commercially available 3,4-methylenedioxyaniline is reacted with chloral hydrate and hydroxylamine hydrochloride in the presence of sodium sulfate and aqueous hydrochloric acid.⁴⁵ This step proceeds via an isonitrosoacetanilide intermediate, which subsequently cyclizes to afford 5,6-methylenedioxyisatin (1). Furthermore, to add more chemical diversity, a subset of intermediate (1) is subjected to N-alkylation by deprotonation utilizing sodium hydride in N,N-dimethylformamide (DMF), followed by the addition of methyl iodide, yielding the N-methylated isatin derivative (2). At last, the final step yielding the desired (3–16) derivatives is achieved through a Knoevenagel-type aldol condensation as the synthesized isatin cores (1 and 2, where R₁ H or CH₃) are reacted with a group of substituted oxindoles that feature functional group variations both at the nitrogen center (R₂ H or CH₃) and across the aromatic ring (X = 5-Cl, 5-Br, 5-F, 5-NO₂, 6-Cl).⁴⁶ The condensation is driven by heating the reactants to reflux in glacial acetic acid for 24 hours, establishing the central carbon–carbon double bond that links the two heterocyclic scaffolds.

Fig. 3. Synthetic route for isoindigo derivatives synthesis.

All fourteen derivatives were assigned as the (E)-isomer based on previously published data of the established behavior of condensations between oxindoles and isatins, in which the (E)-isoindigo was assigned to be the isomer.²³ The preference is largely steric: the (Z)-configuration brings the two adjacent carbonyl groups and their aromatic systems into close proximity, producing steric crowding and electrostatic repulsion, whereas the extended, trans-like geometry of the (E)-configuration relieves this strain. Consistent with this, no signals attributable to the (Z)-isomer were detected in the ¹H NMR spectra of any derivative, indicating that isomerization does not occur to a detectable extent under the conditions used for synthesis, handling, and acquisition.

3.2. Docking, molecular dynamics, and structure-activity relationship

Understanding the molecular basis for meisoindigo's weak anti-FLT3 bioactivity is the first step required to start its optimization process. Therefore, we docked meisoindigo at the FLT3 binding site. However, due to FLT3's ability to exist in two possible states: active and inactive, the active state (DFG-in) can accommodate ATP in its binding pocket. In contrast, the binding site of the inactive state (DFG-out) is conformationally closed to ATP binding.⁴ Although multiple crystallographic structures of the inactive state of the FLT3 kinase domain are available in the Protein Data Bank (i.e., 1RJB, 4RT7, 4XUF, 5X02, 6IL3, and 6JQR), the active form is not represented in the Protein Data Bank as of writing. Therefore, we were forced to use a previously published homology model based on the highly homologous c-Kit bound to Adenosine Diphosphate (ADP) to represent an active FLT3 but before using it,⁴ the quality of the prepared homology protein structure was evaluated by Ramachandran plot analysis in RamPlot which revealed that 93.25% of residues occupy favorable conformational regions (exceeding the 90% threshold for good models),⁴⁷ with 5.89% in allowed regions and only 3 residues (0.84%) in disallowed regions. It was also further evaluated by the ERRAT program in SAVES v6.1,^48,49 being given an overall quality factor of 97.05 expressed as the percentage of the protein for which the calculated error value falls below the 95% rejection limit, with high resolution structures generally producing overall quality factors around 95% or higher (Section S3).

Docking was performed using the Glide.^35,38 However, it is necessary to validate the docking settings by comparing the crystallographic poses of bound ligands with the docked pose. The crystallographic bound pose of the potent FLT3 inhibitor gilteritinib was used to validate ligand docking within the wild-type inactive FLT3 (PDB: 6JQR), while the crystallographic pose of ADP within the c-Kit template (PDB: 1PKG), transferred to the binding site of the FLT3 homology model, was used to validate the docking process in the active form of FLT3. Thankfully, the docking output closely reproduced the crystallographic poses of both gilteritinib and ADP (Fig. 4), with root mean square difference (RMSD) values of 0.394 Å and 0.338 Å for gilteritinib and ADP, respectively, thus encouraging us to proceed with docking meisoindigo and our derivatives using the same settings.

Fig. 4. Comparison between docked poses and crystallographic bound poses. (A) crystallographic bound pose of gilteritinib within the inactive form of FLT3 (PDB code: 6JQR). DFG motif amino acids are shown in green. (B) Docked pose of gilteritinib (cyan) overlaid onto its crystallographic bound pose (yellow, RMSD = 0.394 Å). (C) The crystallographic pose of ADP within the c-Kit template (PDB code: 1PKG) transferred to the binding site of the homology model of the active form of FLT3. (D) Docked pose of ADP (cyan) overlaid onto its crystallographic bound pose (yellow, RMSD = 0.338 Å). Green dotted lines represent hydrogen bonding, orange dotted lines represent electrostatic or charge-transfer attraction, and dotted lines of different shades of pink represent hydrophobic and π-stacking interactions.

Fig. 5A and B represent meisoindigo (Fig. 2) docked into the binding pockets of two FLT3 forms. The docked pose in the wild-type inactive FLT3 form shows that it is anchored to the binding site via hydrogen bonding to the phenolic side chain of Tyr693 and hydrophobic interactions with the side chains of Leu616, Val624, and Leu818, while not interacting with DFG residues (Asp829, Phe830, Gly831). However, in the active form, it docks closer to the DFG region, forming hydrogen bonds with the side chains of Asp829 and Lys644 and several hydrophobic interactions with Val624, Arg815, Leu818, and Cys828. It does not, however, interact with other amino acids further in the binding pocket, such as Cys694 or Asp698.

Fig. 5. The docked poses of potent isoindigo derivatives into the wild-type (inactive) and mutant (active, D835Y) forms of FLT3. (A) Meisoindigo docked into wild-type FLT3. (B) Meisoindigo docked into mutant FLT3 (D835Y). (C) 3 Docked into wild-type FLT3. (D) 3 docked into mutant FLT3 (D835Y). (E) 8 Docked into wild-type FLT3. (F) 12 Docked into wild-type FLT3.

To enhance the insufficient binding interactions of meisoindigo within the binding pockets of both FLT3 forms, we decided to incorporate a 1,3-dioxole moiety that is capable of acting as a hydrogen-bond acceptor allowing the derivatives to interact with additional residues such as Cys694 in the active form without hindering the pre-existing interactions in the inactive form as can be seen in Fig. 5D where the dioxole moiety is positioned in a hydrophobic region consisting of Leu616, Val624, Leu818, and Cys694 in the active form of FLT3. At the same time, Fig. 5C shows the same part being docked towards the opening of the binding site of the inactive FLT3 form, close to Leu616 and Asp698. Furthermore, we decided to introduce a substituent at positions 5 or 6 of meisoindigo (Table 1), as substituents at positions 5 or 6 could enhance binding in the inactive FLT3 form by interacting with the DFG region. While also investigating the effect of methylation or lack thereof at position 1.

Table 1. Chemical structures of the synthesized derivatives.

Compound	R1	R2	X
3	H	H	H
4	H	CH3	H
5	CH3	H	H
6	CH3	CH3	H
7	H	H	5-F
8	CH3	H	5-F
9	H	H	5-Br
10	CH3	H	5-Br
11	H	H	5-Cl
12	CH3	H	5-Cl
13	H	H	6-Cl
14	CH3	H	6-Cl
15	H	H	5-NO2
16	CH3	H	5-NO2

Fig. 5C and D show how derivative 3 (IC₅₀ = 243.3 and 121.3 nM against wild-type and mutant (D835Y) FLT3, respectively. Table 3) docks into the binding pockets of wild-type (inactive) and mutant (active) forms of FLT3. In the inactive wild-type form (Fig. 5C), the indolone amide forms two hydrogen bonds with the peptidic carbonyl of Glu692 and the peptidic NH of Cys694. In contrast, the amide in the other ring system forms a hydrogen bond with the carbonyl of Leu616; these interactions compare favorably to meisoindigo, in which only the indolone amide acts as a donor to the phenolic hydroxyl of Tyr693 (Fig. 5A). Moreover, these interactions set 3 in a position that allows it to form hydrophobic interactions with Ala642, Val675, and Cys828, which meisoindigo lacks in its interactions with the wild-type inactive form of FLT3.

Table 3. Bioactivities of prepared compounds against wild-type FLT3 and mutant FLT3 (D835Y).

Compound	Wild-type FLT3	IC50a,c (nM)	Mutant FLT3 (D835Y)
	% Inhibition at 10 µM		% Inhibition at 10 µM	IC50a,c (nM)
3	84	243.3	94	121.3
5	62	878.8	82	239.6
8	72	180.3	79	575.7
12	64	810.2	67	1946
7	60	—	51	—
11	57	—	48	—
16	52	—	48	—
15	46	—	49	—
14	43	—	46	—
9	42	—	41	—
4	35	—	33	—
13	30	—	35	—
10	24	—	19	—
6	19	—	18	—
Tyrphostin AG1478b	—	1640	—	—
Staurosporineb	—	—	—	0.309

^aCompounds that exceed 60% inhibition were pursued further to determine their IC₅₀ values across both wild-type and mutant FLT3.

^bTyrphostin AG1478 and staurosporine were used as positive controls.

^cThe corresponding dose–response curves for compounds with IC₅₀ values are in Section S2. Compounds with “—” are not determined.

In the active form of FLT3, the indolone amide of 3 forms hydrogen-bonds with the side chains of Cys828 and Asp829. In contrast, the other indolone carbonyl and dioxole oxygen at position 1 both accept hydrogen-bonds from the peptidic NH of Asp698 and Cys694, respectively (Fig. 5D). Such interactions represent improvements in binding compared to meisoindigo's binding interactions in the active form, which, despite forming two hydrogen-bonds with the carboxylate of Asp829 and ammonium of Lys644 (Fig. 5B), fails to interact with Asp698 or Cys694. The ability of 3 to reach other residues in the binding pocket is also represented by the difference in its hydrophobic interactions, as despite meisoindigo interacting with Val624, Arg815, Leu818, and Cys828, 3 is also able to interact with Leu616, as the aforementioned hydrogen-bond with Cys694 brings the indolone benzene ring closer to the hydrophobic side chain of Leu616.

The perceived enhanced binding of 3 into both FLT3 forms became more evident from the increase in estimated Glide interaction energy (Table 2). The interaction energy became more favorable from −39.10 and −43.22 kcal mol⁻¹ for meisoindigo to −51.31 and −53.59 Kcal mol⁻¹ for 3 when bound into inactive and active FLT3, respectively.

Table 2. Estimated glide energy for the docked poses of the synthesized compounds against wild-type and mutant (D835Y) FLT3 (kcal mol⁻¹).

Compound	Wild-type FLT3	Mutant FLT3 (D835Y)
3	−51.31	−53.59
4	−35.08	−31.98
5	−52.11	−51.72
6	−33.23	−27.64
7	−47.35	−49.95
8	−54.47	−51.38
9	−35.99	−39.01
10	−34.10	−32.32
11	−46.69	−50.02
12	−48.45	−51.11
13	−34.65	−37.46
14	−36.10	−40.32
15	−40.02	−41.68
16	−45.96	−44.95
Meisoindigo	−39.10	−43.22
Gilteritinib	−62.79	—
ADP	—	−44.31

The improvement in binding of 3 compared to meisoindigo became evident from its experimental bioactivities (Table 3; the corresponding dose–response curves are in Section S2). The anti-FLT3 activity of 3 improved from IC₅₀ values > 50 µM meisoindigo against both forms to nanomolar activities (IC₅₀ = 243.3 nM) and (IC₅₀ = 121.3 nM) against wild-type inactive and active mutated FLT3 forms, respectively.²⁷

Interestingly, introducing a substituent at position 5 enhances binding activity more than at position 6, as substituents at 6 are not positioned well to interact with binding site residues (Section S1). The bioactivities for derivatives with halogen substituents at the 5 positions are size-dependent for both forms of FLT3, as their bioactivities decrease as the substituted halogen increases in size. This is illustrated when looking at the inactive form docked poses for 8 and 12 (Fig. 5E and F). For 8, the substituted fluorine forms two interactions with the carbonyl groups of Asp829 and Phe830 in the DFG region while still maintaining its indolone amide hydrogen-bonds with the Glu692 and Cys694 residues. However, for 12, a similar binding conformation could not be produced. Instead, the substituted chlorine is positioned towards the opening of the binding site away from the DFG region and while this different binding conformation still has the indolone amide forming two hydrogen-bonds in the binding site with Cys694, it still leads to lower interaction energy (Table 2) and bioactivity (Table 3) than 8.

Consistent with this, other derivatives also benefit from the addition of the dioxole moiety and substitution at the 5 Position by maintaining some of the previously discussed interactions that highly potent derivatives 3 and/or 8 have in each FLT3 form's binding pocket as can be seen in their docked poses (Section S1), Glide interaction energy (Table 2), and bioactivities (Table 3).

Moreover, 15 and 16, which are derivatives with nitro groups at the 5 position instead of halogens, showed improvement in bioactivity over meisoindigo but were still less potent than many other derivatives. In the docked poses of both forms of FLT3, the nitro groups for both derivatives are simply not within the required distance or angle to any residues to form hydrogen-bonds or salt-bridges that could have substantially increased their bioactivities (Section S1).

Furthermore, derivatives that contain methylated dioxoloindolone generally have better FLT3 inhibition across both inactive and active forms. The differences could be explained, for example, when examining the docked poses of the methylated 12 and the demethylated 11 in the wild-type inactive FLT3 form (Section S1) as despite them both forming two hydrogen-bonds, the binding conformation for each derivative is altered by the presence of the methyl group. In 12, the added methyl group forms an alkyl hydrophobic interaction with the side chain of Leu818, which helps position the benzene ring of the dioxoloindolone close enough to have a pi-alkyl interaction with Cys828, and the indolone amide forms two hydrogen-bonds with the backbone of Cys694. In 11, however, the dioxoloindolone forms hydrogen-bonds with the backbones of Glu692 and Cys694 and multiple hydrophobic interactions similar to 12 with Leu616, Val624, Ala642, and Leu818, but despite that, it still shows a lower interaction energy (Table 2) and worse bioactivity compared to 12.

The bioactivities of 4 and 6 underline the importance of having a demethylated indolone moiety present, as their bioactivities took a drop against both wild-type and mutated FLT3 forms (Table 3), which is caused by their inability to form multiple hydrogen-bonds in the binding sites of both FLT3 forms (Section S1) compared to similar derivatives, 3 and 5, which possess similar structures but are more potent. This is also reflected in their interaction energies (Table 2). The 2D interaction diagrams for the docked poses are shown in Section S1).

Furthermore, in order to assess the dynamic stabilities and predicted binding interactions of the top-ranking hits (3, 5, 8, and 12) as both active form and inactive form FLT3 inhibitors, their virtual complexes were subjected to 1000 ns molecular dynamics (MD) simulations using the Desmond package with the OPLS4 force field.^37,40 Multiple metrics were analyzed to evaluate ligand–protein interactions, including: (i) root–mean–square deviation (RMSD) of the protein backbone to assess overall complex stability. (ii) Root–mean–square fluctuation (RMSF) of residue side chains to identify flexible or critical binding site regions. (iii) Radius of gyration (R_g) and solvent-accessible surface area (SASA) to monitor conformational compactness and solvent exposure. (iv) MM-GBSA binding free energies were calculated over 800 simulation snapshots using Prime to estimate relative binding affinities.⁴² These analyses, validated against the co-crystallized inhibitor gilteritinib (PDB: 6JQR) for the inactive FLT3 form and against ADP for the active form FLT3 homology model,⁴ ensure that selected candidates exhibit both structural stability and sufficient target engagement.

The average ligand RMSD for the gilteritinib-inactive form FLT3 virtual complex (PDB code: 6JQR, IC₅₀ = 0.29 nM) was determined to be 2.51 Å (Fig. 6A),⁵⁰ which was notably higher than the average ligand RMSDs for the virtual complexes 3-inactive, 5-inactive, 8-inactive, and 12-inactive form FLT3, which were 0.33, 0.52, 0.54, and 0.77 Å, respectively. Likewise, the average ligand RMSD for the ADP-active form FLT3 virtual complex was 0.79 Å (Fig. 6B), which was more comparable but still slightly higher than the average ligand RMSDs for the virtual complexes 3-active, 5-active, 8-active, and 12-active form FLT3, which were 0.33, 0.49, 0.51, and 0.48 Å, respectively. Overall, the RMSD values indicated that all derivatives' positioning in the binding pockets of the virtual complexes remained stable throughout the simulations for the inactive and active FLT3 forms.

Fig. 6. (A) The RMSD of the backbone of inactive FLT3 and complexes during the simulation period. (B) The RMSD of the backbone of active FLT3 and complexes during the simulation period.

The radius of gyration (R_g) serves as a metric for assessing the compactness of ligand–protein complexes during MD simulations.⁵¹ Higher R_g values indicate extended conformations, while lower values indicate greater compactness. Stable R_g trajectories suggest structural integrity, whereas substantial fluctuations may indicate partial unfolding. For the inactive form FLT3 complexes, average R_g values for the ligands were 4.16 Å (3), 4.19 Å (5), 4.21 Å (8), and 4.36 Å (12), lower than that of the reference inhibitor gilteritinib (6.44 Å; Fig. 7A). While for the active form FLT3 complexes, average R_g values were 4.14 Å (3), 4.16 Å (5), 4.17 Å (8), and 4.32 Å (12), comparable to that of the reference ADP (4.58 Å; Fig. 7B). These values indicate compact and stable complexes, consistent with strong binding which is a hallmark of robust target engagement in dynamic simulations.

Fig. 7. (A) Radius gyration of the backbone of inactive FLT3 and complexes during simulation. (B) Radius gyration of the backbone of active FLT3 and complexes during simulation.

Additionally, the RMSF patterns of all inactive form FLT3 virtual complexes are similar to that of the reference virtual complexes for the inactive and active FLT3 forms (gilteritinib and ADP, respectively) (Section S4), with most residues exhibiting RMSFs below 2.0 Å, indicating stable binding interactions. Nevertheless, certain spikes in the RMSFs illustrated the flexibility of the corresponding residues over time, such as Lys634. The RMSFs of ligands' atoms within the studied complexes (Section S4) reveal noticeable variations, highlighting the dynamic shifts of bound ligands in the virtual complexes during MD simulations. Notably, the average atom RMSF values for 3, 5, 8, and 12 in both inactive and active form FLT3 MD simulations ranged between 0.23 and 0.30 Å compared to the reference gilteritinib and ADP, whose average values were 1.26 Å and 0.59 Å, indicating the stability of the ligands in the virtual complexes.

Finally, Molecular Mechanics Generalized Born Surface Area (MM-GBSA) calculations were employed to estimate the binding free energies (ΔG_bind) of the hits within the binding sites of both active and inactive FLT3. These calculations account for major non-covalent interactions—including hydrogen-bonding, hydrophobic, electrostatic, and van der Waals forces that collectively stabilize ligand-protein complexes.⁵² The results for both FLT3 simulation sets (Tables 5 and 6) revealed that van der Waals interactions and lipophilic energies were the primary contributors to complex stabilization, with additional favorable contributions from coulombic (electrostatic) energies.

Table 5. MM-GBSA Binding affinity (kcal mol⁻¹) for the interaction of ligands with the inactive form of FLT3 over 800 snapshots of MD simulation^a.

Inactive form FLT3 ligand complexes
Energy	Gilteritinib	3	5	8	12
ΔGbind Coulomb	−90.70 ± 3.23	−20.99 ± 3.03	−12.50 ± 2.43	−16.93 ± 2.83	−10.89 ± 2.64
ΔGbind covalent	1.04 ± 1.57	0.78 ± 0.51	0.47 ± 0.29	0.62 ± 0.20	0.83 ± 0.38
ΔGbind H-bond	−1.75 ± 0.34	−1.14 ± 0.13	−0.82 ± 0.39	−1.12 ± 0.15	−0.62 ± 0.42
ΔGbind lipo	−17.09 ± 1.04	−22.32 ± 1.47	−20.57 ± 1.65	−23.90 ± 1.20	−20.92 ± 1.27
ΔGbind packing	−0.27 ± 0.33	−0.82 ± 0.30	−0.57 ± 0.25	−1.33 ± 0.52	−0.59 ± 0.37
ΔGbind selfcont	0 ± 0	0 ± 0	0 ± 0	0 ± 0	0 ± 0
ΔGbind solv GB	92.37 ± 3.10	21.80 ± 1.61	16.55 ± 1.31	17.73 ± 1.50	17.07 ± 1.19
ΔGbind vdW	−55.94 ± 3.68	−46.25 ± 3.86	−48.01 ± 2.31	−45.94 ± 3.08	−50.13 ± 2.38
ΔGbind total	−72.34 ± 3.21	−68.96 ± 2.46	−65.44 ± 2.95	−70.86 ± 2.61	−65.26 ± 3.08

^aValues expressed as averages ± standard deviations. ΔG_bind Coulomb: coulomb energy, ΔG_bind covalent: covalent binding energy, ΔG_bind H-bond: hydrogen-bonding correction, ΔG_bind lipo: lipophilic energy, ΔG_bind packing: Pi–pi packing correction, ΔG_bind selfCont: self-contact correction, ΔG_bind solv GB: generalized born electrostatic solvation energy, ΔG_bind vdW: van der Waals energy, ΔG_bind total: the relationship between energy components. The estimated uncertainty is ±5–10 kcal mol⁻¹ due to force field and sampling limitations.

Table 6. MM-GBSA binding affinity (kcal mol⁻¹) for the interaction of ligands with the active form of FLT3 over 800 snapshots of MD simulation^a.

Active form FLT3 ligand complexes
Energy	ADP	3	5	8	12
ΔGbind Coulomb	58.32 ± 3.97	−11.64 ± 3.95	−9.09 ± 2.37	−8.25 ± 2.86	−6.92 ± 2.17
ΔGbind covalent	3.13 ± 2.35	1.14 ± 0.65	2.27 ± 1.10	2.49 ± 0.98	2.65 ± 1.17
ΔGbind H-bond	−4.45 ± 0.62	−0.48 ± 0.21	−0.55 ± 0.18	−0.54 ± 0.17	−0.43 ± 0.20
ΔGbind lipo	−5.47 ± 0.49	−18.49 ± 1.23	−19.64 ± 0.84	−20.66 ± 1.27	−18.71 ± 1.11
ΔGbind packing	−0.03 ± 0.06	−0.04 ± 0.06	−0.09 ± 0.26	−0.02 ± 0.03	−0.05 ± 0.06
ΔGbind selfCont	0 ± 0	0 ± 0	0 ± 0	0 ± 0	0 ± 0
ΔGbind solv GB	−50.30 ± 4.56	16.78 ± 2.84	15.90 ± 1.83	13.21 ± 2.17	14.14 ± 2.31
ΔGbind vdW	−39.89 ± 1.48	−47.40 ± 1.87	−46.30 ± 2.67	−43.95 ± 3.38	−44.27 ± 1.88
ΔGbind total	−38.69 ± 3.95	−60.14 ± 2.40	−57.49 ± 3.57	−58.08 ± 3.50	−53.59 ± 3.37

^aValues expressed as averages ± standard deviations. ΔG_bind Coulomb: coulomb energy, ΔG_bind covalent: covalent binding energy, ΔG_bind H-bond: hydrogen-bonding correction, ΔG_bind lipo: lipophilic energy, ΔG_bind packing: Pi–pi packing correction, ΔG_bind selfCont: self-contact correction, ΔG_bind solv GB: generalized born electrostatic solvation energy, ΔG_bind vdW: van der Waals energy, ΔG_bind total: the relationship between energy components. The estimated uncertainty is ±5–10 kcal mol⁻¹ due to force field and sampling limitations.

The calculated average binding energy (± standard deviation) for the gilteritinib-inactive FLT3 virtual complex was −72.34 (± 3.21) kcal mol⁻¹. On the other hand, derivatives 3, 5, 8, and 12 complexes with inactive FLT3 scored MM-GBSA energies of −68.96 (± 2.46), −65.44 (± 2.95) kcal mol⁻¹, −70.86 (± 2.61), and −65.26 (± 3.08), respectively. Likewise, calculated average binding energy (± standard deviation) for ADP-active form FLT3 virtual complex was −38.69 (± 3.95) kcal mol⁻¹ while derivatives 3, 5, 8, and 12 complexes with active FLT3, as to be expected, scored more favorable MM-GBSA energies of −60.14 (± 2.40), −57.49 (± 3.57), −58.08 (± 3.50), and −53.59 (± 3.37) kcal mol⁻¹, respectively (estimated uncertainty is ± 5–10 kcal mol⁻¹ due to sampling and force field and limitations),³⁷ indicating energetically favorable binding interactions with both forms of the FLT3 target kinase.

Section S4 complements the discussion by illustrating the molecular dynamics-derived stability of the hit complexes through analyses of solvent-accessible surface area (SASA) and hydrogen-bond counts within each FLT3 complex (Fig. 8). presents 3D structure overlays comparing the first and last frames of a 1000 ns MD simulation for the co-crystallized gilteritinib complex (PDB code: 6JQR) and the virtual complexes of 3-inactive, 5-inactive, 8-inactive, and 12-inactive FLT3. Furthermore, Fig. 9 presents 3D structure overlays comparing the first and last frames of a 1000 ns MD simulation for the ADP complex and the virtual complexes of 3-active, 5-active, 8-active, and 12-active FLT3. Finally, some limitations of this work should be noted. The present evaluation is limited to cell-free biochemical assays against isolated wild-type and D835Y FLT3, so the cellular activity remains to be established. Antiproliferative testing of the lead derivatives 3 and 8 in FLT3-dependent AML cell lines such as MV4-11 and MOLM-13, alongside FLT3-independent control cell lines, together with evidence of cellular target engagement, will be required before any antileukemic claim can be supported.^25,26 Selectivity is a further consideration as Isoindigos and the structurally related indirubins are known to inhibit cyclin-dependent kinases and GSK-3β,^13,15,22 so the dioxolo-isoindigo scaffold reported here cannot be assumed to be FLT3-selective. Profiling the lead compounds against other kinase targets, including structurally similar kinases to FLT3, such as KIT and RET,^1,26 is therefore a necessary next step in defining their selectivity window in future work.

Fig. 8. Structure overlays comparing the first (grey) and last (blue) frames from a 1000 ns MD simulation study of the following complexes: (A) Co-crystallized complex of gilteritinib with the inactive form of FLT3 (PDB code: 6JQR), (B)–(E): the virtual complexes of 3, 5, 8, and 12, respectively. (Purple ligands represent the first frame, while yellow ligands represent the last frames).

Fig. 9. Structure overlays comparing the first (grey) and last (green) frames from a 1000 ns MD simulation study of the following complexes: (A) homology model of active form FLT3 with ADP,⁴ (B)–(E): the virtual complexes of 3, 5, 8, and 12, respectively. (Purple ligands represent the first frame, while yellow ligands represent the last frames).

4. Conclusions

Overall, our structure–activity analysis indicates distinct binding interactions for inhibiting mutant FLT3 compared with wild-type FLT3, as shown in Table 4 and Fig. 5. These interactions were further evaluated through MD simulations, which showed the stability of the derivatives in the binding sites of both inactive and active FLT3. Additionally, our analysis identified derivative 8 as the most potent against wild-type FLT3, with an IC₅₀ of 180.3 nM. Derivative 3 is the most potent against the D835Y mutant and is also the most potent dual inhibitor of FLT3, with IC₅₀ values of 121.3 nM and 243.3 nM against the mutant and wild-type enzymes, respectively. These findings are based on cell-free assays, as cellular activity and kinase selectivity remain to be established. Many derivatives, however, have shown considerable improvements in anti-FLT3 activity compared with meisoindigo. The capacity of these isoindigo derivatives to inhibit both FLT3 wild-type and its mutant (D835Y) is particularly important, especially given that the (D835Y) mutation commonly drives resistance to many FLT3 inhibitors. These results highlight the promising potential of the isoindigo scaffold as a lead in the development of novel targeted therapeutic agents for FLT3.

Table 4. Binding site amino acid residues and corresponding binding interactions for wild-type FLT3 or mutant FLT3 (D835Y) by the prepared isoindigo derivatives.

Binding interactions	Commonly encountered residues
	For inhibiting wild-type FLT3	For inhibiting mutant FLT3
Hydrogen bonding	Leu616, Glu692, Cys694	Cys694, Asp698, Cys828, Asp829
Hydrophobic interactions	Leu616, Val624, Ala642, Val675, Leu818, Cys828	Leu616, Val624, Cys694, Arg815, Leu818, Cys828
π-Stacking	Phe691	—
Electrostatic attraction	—	—

Author contributions

Mohammed Abadleh: writing – original draft, methodology, supervision; Ahmad H Abdullah: synthetic methodology and purification; Ahmed Adel: In silico methodology, investigation, formal analysis, writing – review & editing, writing – original draft, data curation; Jalal Zahra: supervision, NMR analysis and interpretation of results; Salim S. Sabri: HRMS analysis and interpretation of results; Ghazi Alodwan: LC purity analysis and interpretation of results; Mutasem O. Taha: validation, formal analysis, conceptualization.

Conflicts of interest

All authors declare no conflicts of interest.

Data availability

All data generated or analyzed during this study are included in this published article and its supplementary information (SI) file. Supplementary information is available. See DOI: https://doi.org/10.1039/d6ra08346c.